How to appear in ChatGPT and show up in AI search

To appear in ChatGPT and other AI search results, start with useful pages that search services can crawl and index. Answer real customer questions clearly, keep business details accurate and check what each engine actually returns. Search access and model training use different controls. An optional llms.txt can help supported tools discover documentation; it is not a universal ranking signal or a requirement for Google AI features.

First, understand what you're actually optimising for

Old search hands you ten blue links and you pick one. ChatGPT and Perplexity read a pile of pages, write a paragraph, and name a couple of sources. The prize isn't a ranking any more; it's getting quoted. That's the whole shift, and it changes the job. You're not trying to be the tenth-best page on a topic so you scrape onto page one; you're trying to have the one clean, liftable sentence that answers the exact question, so the engine grabs yours.

This has a name now: answer-engine optimisation, or AEO, and if you want the acronyms untangled we did that in AEO, GEO and AIO explained. For why this is worth your attention this year rather than next, read why AI search matters now. Ignore the jargon. It's the same instinct good tradespeople have always had: say the useful thing first, plainly, and don't make people dig for it. The machines just reward it more obviously than Google ever did.

Choose search access and training separately

Check robots.txt as well as your CDN, firewall and server responses. Allowing a search crawler makes access possible; it does not promise indexing or a citation. A training crawler has a different purpose.

ProviderSearchUser-requested accessModel training
OpenAIOAI-SearchBotChatGPT-UserGPTBot
AnthropicClaude-SearchBotClaude-UserClaudeBot
GoogleGooglebot controls Search eligibility, including AI featuresProduct-specific retrieval controls varyGoogle-Extended controls specified Gemini and Vertex AI uses, not Google Search inclusion

For example, OpenAI documents separate search and training controls. User-triggered requests can behave differently from automated crawling, so read each provider's rules rather than treating every user agent alike. See OpenAI's crawler purposes, Anthropic's crawler controls and Google's AI feature guidance. Perplexity publishes its own crawler roles.

Write the answer so a machine can lift it

This is where the real work is, and it's writing, not tooling. An engine doesn't quote your page; it quotes a passage from it. So every page that targets a question should answer that question plainly, near the top, in a paragraph that stands on its own without the three paragraphs of throat-clearing above it.

Practically, that means:

  1. Lead with the answer, then back it up. Put the verdict in the first hundred words. "A small brochure site sits in the low thousands" beats "pricing depends on many factors" every single time, because the first one is quotable and the second one is noise.
  2. Use headings that match how people ask. Real questions as your H2s ("How much does it cost?", "Do I need one?"), not cute marketing labels. The engine maps the question to your heading.
  3. Keep the core answer tight. Roughly forty to sixty words is the sweet spot for a passage an engine can lift whole. Say the thing, then expand below it for the humans who want the detail.
  4. Add a real FAQ at the bottom. Natural-language questions with straight answers. It's the single most quotable block on a page, which is exactly why this article has one.

The trap to avoid: do not write a separate "AI version" of your content, keyword-stuffed and robotic. Google explicitly calls that scaled content abuse and it can earn you a penalty, not a citation. Write one good page for a human and structure it so a machine can read it too. That's not two jobs; it's one job done properly. Keyword stuffing actively goes backwards here; the engines are reading for meaning, not counting your repetitions.

llms.txt: an optional discovery file

llms.txt is a proposed Markdown map of important pages, described at llmstxt.org. Some documentation and agent tools support it. That does not establish that every search engine consumes it or uses it to recommend a business. Google says no special AI text file is needed for its AI features. Keep important information in accessible HTML first; add this optional file where a documented consumer makes it useful.

# Your Business Name
> One plain line on what you do and who you do it for.

## Key pages
- [Services](https://yourdomain.com.au/services.html): what you offer
- [About](https://yourdomain.com.au/about.html): who you are
- [Contact](https://yourdomain.com.au/contact.html): how to reach you

If a tool you support consumes Markdown, you can also provide a maintained pricing file. Keep it consistent with the visible prices. An additional file should not become a second, conflicting version of the offer; visible, accessible pricing is the priority.

IndexNow: the push lever people install and never pull

Here's a lever most sites leave sitting idle. Several of the AI answer engines lean on Bing-family indexes to decide what to quote, and IndexNow is the way you actively push a changed page at that index instead of waiting to be crawled. It's the difference between knocking on the door and hoping someone walks past your window.

The catch that trips everyone up: having the IndexNow key file sitting on your server is not the same as ever pushing anything. The key just proves the submissions are yours. If nothing actually sends the ping when a page changes, the file is decoration and your updates crawl in on the slow path like everyone else's. Plenty of sites tick "IndexNow enabled" and never fire a single submission.

So verify the submissions genuinely run, not just that the plugin box is ticked, and treat a resubmit as part of publishing. Every time you change real content, resubmit your sitemap and fire an IndexNow ping for the changed URLs. Do it after every edit, not only on launch day, because the whole point of a push lever is to use it when the thing you want indexed actually changes.

The part nobody selling "AI SEO" wants to admit

Half your AI visibility isn't on your website at all. These engines cite where you appear (a local directory, a review site, an industry roundup, a Reddit thread), often ahead of your own pages, because a third party vouching for you reads as more trustworthy than you vouching for yourself. So the off-site work matters: accurate listings in the directories that cover your trade, real reviews where your industry actually leaves them, getting named in a "best X in [your town]" piece. That's not a growth hack; it's being genuinely present where people already talk about your kind of work.

And the boring foundation underneath all of it: a fast, clean, properly structured website. AI engines read the same HTML a browser does, so if your content only appears after four frameworks finish loading, the bot sees a blank page. Semantic HTML, real headings, content that's actually in the page: the fundamentals that make a site rank and load well are the same ones that make it readable to a machine. There's no separate AI website; there's a good website, and AI rewards it. It's the groundwork the whole cluster of get-found guides sits on.

The honest order to do this in

Don't boil the ocean. Work down this list and you'll have done more than nearly every competitor in your area:

  1. Unblock the bots. Check robots.txt, allow the search crawlers in the table above. Free, five minutes, and the one that stops you being invisible.
  2. Fix your top three pages. Lead each with a plain answer, question-shaped headings, a real FAQ at the bottom. This is the work that gets you quoted.
  3. Consider optional discovery files last. Use llms.txt or Markdown for documented consumers and keep them in step with the HTML.
  4. Get present off-site. Accurate directory listings, real reviews, an honest mention or two where your trade gets discussed.
  5. Sort the foundation. If the site's slow or the content doesn't render cleanly, fix that; it's holding back your normal search too.

Notice what's not on the list: paying a monthly "AI visibility" subscription to a tool that promises to "optimise you for ChatGPT." You don't need it to get started, and most of what they'd do is the five steps above. This is the same call as the older question of whether to spend on SEO or Google Ads first: get the free, structural wins in the bag before anyone takes a retainer off you.

An Australian worked example: a Mosman chiropractor

ReFigure Chiropractic is a clinic in Mosman, on Sydney's Lower North Shore. Before the rebuild it had 133 five-star Google reviews and two Google impressions in 90 days. In August 2026, Google's AI answer for "who is the best chiro in mosman" named ReFigure first and cited the clinic's own pages.

Four things made that answer possible:

  1. Search engines could reach the pages again. The old domain forwarded page for page, and a certificate fault that stopped Google from following those forwards was fixed.
  2. Each page answered a real question plainly. The copy was also checked against AHPRA advertising rules, and sixteen items that implied a treatment outcome were removed.
  3. The facts were in structured data. The organisation, services, FAQs, practitioners and reviews were marked up on every page, so an engine could quote the clinic accurately.
  4. The off-site proof was already there. The 133 reviews were the evidence. The site made the clinic easy to find and to read.

Measured in the clinic's Search Console: 3,054 Google impressions in the 44 days from 6 July to 18 August 2026. The clinic also ranked number one for "chiro near me" and switched off more than $800 a month in Google Ads.

What "showing up" actually looks like

The durable work is a useful, accessible site, clear answers supported by real evidence, accurate business information and a trustworthy presence beyond your own domain. Measure actual answers over time. Those foundations make a page easier to discover and understand; no file or format guarantees a recommendation.

Where this comes from

The primary sources behind the above, so you can check any of it yourself rather than take our word for it:

Frequently asked questions

How do I appear in ChatGPT and other AI search engines?

Make useful pages accessible to the relevant search crawlers, answer customer questions clearly and keep business details accurate. Check indexing and actual engine answers. Search and training crawlers have different purposes; allowing GPTBot is not a requirement for ChatGPT search. An optional llms.txt is a discovery aid for supported tools, not a universal AI ranking requirement.

What is llms.txt and do I actually need one?

llms.txt is a proposed Markdown map of useful pages at your domain root. Some documentation and agent tools support it. It is not a universal search-ranking signal, and Google does not require it for AI features. Keep your HTML useful and crawlable first; maintain the optional file when a documented consumer needs it.

Will optimising for AI search hurt my normal Google ranking?

No, because the work is the same work. Google's own line is that its AI answers run on its normal search ranking, so good, honest, clearly-structured content is what wins in both. The mistake to avoid is writing a separate, robotic "version for AI" stuffed with keywords. Google calls that scaled content abuse and it can get you penalised. Write one good page for humans, structured so a machine can read it too, and you've covered both at once.

Should I block AI bots to protect my content?

Decide search access and training separately. OpenAI distinguishes OAI-SearchBot for search from GPTBot for training; Anthropic distinguishes Claude-SearchBot from ClaudeBot. User-triggered retrieval has its own rules. Check provider documentation and firewall behaviour before changing access. Blocking a training crawler does not automatically block the same provider's search crawler.

Why does AI quote a competitor and not me?

Usually one of three reasons. Either your site blocks the bots, so you're invisible to start with. Or your page never plainly answers the question, so there's nothing clean to quote, while the competitor has a tidy paragraph that does. Or the engine is pulling from somewhere you're not, like a directory, a review site or a Reddit thread, because AI cites where you appear, not just your own site. Fix the first two on your page, then go get mentioned in the places the engines already trust.

Want this done rather than fiddled with? We build small-business sites that are fast, clean and structured to be found (by Google and by the AI engines), and we'll set up your robots.txt, llms.txt and the answer-shaped pages as part of the job, no monthly "AI" subscription bolted on. It's also part of our get-online bundle for new businesses who want the lot sorted in one go. Tell us what you do and who you do it for, and we'll tell you straight where your first effort is best spent.